Training an Agent to Find and Reach an Object in Different Environments using Visual Reinforcement Learning and Transfer Learning.
Evelyn Conceição Santos BatistaWouter CaarlsLeonardo A. ForeroMarco Aurélio PachecoPublished in: ICAART (2) (2021)
Keyphrases
- transfer learning
- reinforcement learning
- multi agent environments
- reinforcement learning agents
- multi agent
- learning tasks
- dynamic environments
- action selection
- training and test data
- knowledge transfer
- supervised learning
- learning agent
- cross domain
- active learning
- labeled data
- reward function
- semi supervised learning
- transfer knowledge
- machine learning
- domain adaptation
- state abstraction
- multi task
- transferring knowledge
- learning algorithm
- multi task learning
- model free
- collaborative filtering
- reinforcement learning algorithms
- markov decision processes
- manifold alignment
- machine learning algorithms
- training examples
- state space
- real world
- action space
- semi supervised
- target domain
- text classification
- unsupervised learning
- previously learned
- learning process
- training set
- visual features
- learning problems
- information retrieval
- optimal policy